A Deep Learning Approach for Predicting Student Academic Performance and Grades Based on C Code Analysis

Abstract

This research employs deep learning to enhance student assessment by analyzing the quality and structure of programming assignments, focusing on C code submissions. Traditional grading methods often fail to capture the intricate details of a student’s coding abilities, focusing primarily on code functionality over quality and comprehension. To address this limitation, it requires to extracts in-depth metrics from code—such as the total lines of code, use and quality of comments, variable declarations, control structures, and overall readability—and integrates them with traditional academic data like past grades and attendance. Using a Deep Neural network, the model predicts students' grades and academic performance percentages based on these rich, combined inputs, providing a more comprehensive and nuanced evaluation. This innovative approach empowers educators with insights into each student’s overall performance and areas needing improvement, enabling personalized feedback and fostering a balanced, skill-based assessment framework that goes beyond conventional grading systems.